I’m definitely not an epidemiologist, so take this post with thousands of grains of salt. But I am trying to think about whether we should expect a major second wave of COVID-19.

Andrew Atkeson, Karen Kopecky, Tao Zha look at the 23 countries and 25 states with the highest death tolls and find a consistent pattern for all of them. One clear peak has been followed by “relatively slow growth or even shrinkage of daily deaths from the disease.” These are illustrations of the classic pattern:

There is enormous variation in the death rate at the peak. For instance, at their respective peaks, 24 people per million died each day in Belgium, versus 0.27 per million in New Zealand. Yet most states and countries–and all the ones included by Atkeson, Kopecky, and Zha–look similar 20-30 days after the peak. Belgium, for example, has had less than one daily death per million since June 12.

However, some countries and states do not exhibit this pattern. I have found pretty clear evidence of second peaks in Croatia, Iran, Israel, Japan, and Turkey, plus Idaho, Louisiana, and Mississippi.

I included the USA in the graph because it also shows two humps (the second smaller than the first). However, disaggregating US data to the state level suggests that there were simply two batches of states that had one peak each. At the state level, the only true second peaks that I see are in Idaho, Louisiana, and Mississippi. There are also some cases, like Australia, in which you can see a second peak if you squint–but the death rate has never been high. And there are countries, like Ukraine, that seem to wobble upward slowly without peaking,

Reading Atkeson, Kopecky, and Zha, one might guess that most badly-afflicted countries have accomplished impressive declines by implementing interventions. That is not such good news, since these policies are costly and hard to sustain. But it would be surprising if all the jurisdictions in their sample accomplished the same outcome in 20-30 days despite applying divergent policies. There is some chatter that these places have reached herd immunity, but I am convinced by Howard Forman and others that’s not what’s happening. Still there could be a strong tendency for COVID-19 to taper off for other reasons, which might offer good news.

It could also be the case that we simply haven’t seen many second waves yet. When you play Russian Roulette, things often go fine for a while, but the game always ends the same. Possibly places like Turkey and Croatia and Idaho and Louisiana demonstrate that we’re all at risk of a resurgence at a random moment.

Some European countries have recently reported increases in cases, although not deaths. Perhaps this is only because of increased testing rates–but then again, why is testing becoming more common unless rising numbers of people are experiencing symptoms? Deaths may follow.

In any event, I am searching and waiting for more information about the actual second waves. Why have they happened and what can we learn from their experiences?